The Capital Might Be Misaligned, But The Plane is Probably Fine
A response to Mike Brock’s “The Capital is Misaligned and the Crash is Coming” drafted by Opus 4.7 based on a text thread with my friend who sent the article and edited by me.
Really interesting read. I agree with most of the points, but I think the conclusion might be wrong. That said, Brock is clearly smart and I don’t even pretend to understand the market, so feel free to ignore me.
Where I agree
Agree with 100% of his opening point:
The first is that AI is the future. Large language models are the most significant advance in the history of software, and the industry has not yet absorbed what that means. The second is that the current AI buildout is a bubble — not because the technology is overhyped, but because the capital being deployed is massively misaligned with where the technology is actually going. Both claims are true simultaneously. That simultaneity is what neither Wall Street nor the hyperscaler executives are prepared to face, because the implication is that the entire cloud-software business model of the last fifteen years is about to be dismantled by the technology the same industry is pouring trillions into building.
And I strongly agree that people are sleeping on Apple. This part from the article feels right:
Within the next several iterations of the hardware — I am not making a ten-year prediction, I am making a short-horizon prediction — Apple will be running a local Siri on your phone that is as capable as ChatGPT or Claude is today. For free. Without sending your queries to a server. Without training on your data. Without rate limits. Without surveillance.
Yes. Everyone will have a personal assistant / open-Claude-5.0 in their pocket. That seems like a good bet, and it’s probably a good argument to buy some Apple stock.
Where I disagree
But I’m not at all convinced that Mythos + 2 will run on an iPhone.
And Mythos + 2 is what will build all the software.
And that will be on the massive data center infrastructure being built.
So yes, the pocket-assistant future arrives. But the frontier models will still need GPU towers in data centers. The article collapses these two things together, and I don’t think they collapse.
Could it still be a bubble? Sure. Maybe. The open question is whether there’s really a Mythos 3.0 coming with a continued value add — especially if RL or other breakthroughs don’t hit. I suspect there is, but I’m not at all sure about that.
The revenue question
My friend pushed back on this: will it add value for the masses? At what cost? The ultimate question is revenue. I’m hitting my limit on Opus constantly now. And then: if you can’t offer Mythos 3.0 at a reasonable price, does it matter?
Fair question. Here’s my intuition: elite software engineers (but not generic PhDs) running in a data center (but not in an iPhone app) are about a year away. That alone, I think, carries the bubble for quite a while.
Like — I suspect I will want to “hire” a bunch of them.
It’s SO MUCH WORK to “manage” my agents right now. I think that friction should reduce to the point that it essentially replaces SaaS. Mythos seems to take a huge step in that direction.
So people (e.g. me) pay for “agents” instead of SaaS. That’s the revenue.
And SaaS is nothing close to reasonably priced now, IMHO. So there’s a relatively low bar for Anthropic / GPT and friends to replace it. That’s the bar they have to cross — and they’re doing it easily now. The next batches of use cases, though, get more and more complex.
My friends’s counter: the revenue implied just by OpenAI and Anthropic investments is an order of magnitude higher than basically all SaaS combined.
Yeah. Which is why there still might be a bubble. But…
We might be at the beginning of software
I’ve read a few articles about the idea that “we’re at the beginning of software.” Basically: there’s 10x or 100x or 1000x more software that will be created in the coming years. I might agree. There’s at least 10x, maybe less than 1000x, in clinical trials (my world). There’s just so much that seems like it’s possible now — or at least in the next few years — that wasn’t worth considering before.
I’m personally considering a bunch of things that just weren’t an option before, but become pretty reasonable if I can “hire” 100x engineers at 0.01 the cost.
The question is really just about how much the friction goes down. There’s no way in hell I can manage 100 agents now — especially in my highly regulated environment where their decisions affect people’s health. But it doesn’t feel too crazy that it will be doable in a year or two. Not a lock. Not impossible either.
I’m really curious about how much Mythos unlocks on this front. Someone really needs to create the “Agentic Technical PM” that a friend and I were discussing at dinner. That unlocks quite a lot, I think.
So, bubble or not?
Even with all of that, I wouldn’t rule out a bubble. There’s some chance of significant progress toward AGI — maybe those extra layers of software unlock something, maybe there’s some other breakthrough, maybe the LLM exponential goes further than some people (myself included) think.
Just really hard to say. Huge variability. Incredibly hard to predict.
Brock’s core claim — the technology is real and the capital is misaligned — is internally coherent, and it’s the kind of claim that could be right. But it rests on two assumptions I’d push on: that local inference eats the frontier, and that the revenue thesis is just “replacing SaaS.” I don’t think local inference eats the frontier, and I think the revenue thesis is bigger than SaaS replacement — it’s every piece of software that was too expensive to build before.
Maybe that still doesn’t close the gap to current valuations. Maybe it does. I genuinely don’t know. But I don’t think the case is as settled as the article makes it sound.